OpenAI announces GPT-5 upgrade: Provides "automatic", "fast", and "deep thinking" modes, personality becomes more affable

Phoenix News Tech Report August 13, OpenAI CEO Otterman published a dynamic on X platform, detailing the latest updates to ChatGPT. This time, the core update is GPT-5's major breakthroughs in multiple key areas.

In response mode, GPT-5 provides three options: "automatic", "fast", and "deep thinking". Otterman notes that the "automatic" mode is suitable for most users, which can intelligently match the best answer method. The new control options provide more convenience for those with specific requirements, meeting their needs for answering speed or depth in different scenarios.

In terms of personality shaping, the upcoming GPT-5 version will become even more affable. This improvement is based on feedback from most users, aiming to avoid the overenthusiasm issues that occurred with GPT-4. Otterman emphasizes that the team deeply recognizes the importance of implementing personalized AI personality customization functions in the future, striving to make user-AI interactions more natural and comfortable.

According to reports, GPT-5 employs an embedded three-in-one integrated architecture composed of the GPT-5-main model for handling regular questions, the GPT-5-thinking deep thinking model for solving complex tasks, and a real-time decision-making mechanism. When the quota is used up, it will start a mini version. It can quickly decide which model to use based on dialogue type, complexity, tool requirements, and user intentions, and autonomously judge whether to enter deep thinking mode, thereby automatically matching the most suitable model to answer questions.

According to OpenAI's publicly released test data, GPT-5 has achieved significant performance improvements in mathematics, programming, visual perception, and health domains compared to previous models.

Currently, GPT-5 can be used for free, with higher intelligence levels requiring a paid upgrade to Plus or Pro users.